NumPy: Creating Arrays

Q: What's the modern way to generate random numbers?

Use np.random.default_rng(seed) to create a Generator object, then call .random() , .normal() , .integers() , etc. The old np.random.seed() still works but is considered legacy.

Q: Can I create an array from a generator expression?

np.array(x for x in range(10)) doesn't work — it creates a 0D array containing the generator object. Use np.fromiter((x for x in range(10)), dtype=int) instead.

Q: How do I handle missing data when loading CSV?

Use np.genfromtxt('file.csv', delimiter=',', filling_values=0) to replace missing values with a default, or np.nan .

Last updated: 2026-08-26

Creating Arrays

1. What You'll Learn

❶ Creating from Python objects
❷ Numeric sequences (arange / linspace / logspace)
❸ Pre-filled arrays (zeros / ones / eye / full)
❹ Random number generation
❺ Reading from files

2. Story

Charlie needs 100 evenly spaced angles from 0 to 2π. Python loop: 5 lines. np.linspace: 1 line.

linspace(0, 2*pi, 100) means "take 100 evenly spaced points from 0 to 2π" — code is documentation.

3. Key Concepts

(1) Creating from Python Objects

The most basic approach — converting Python lists (or nested lists) to NumPy arrays:

PYTHON

import numpy as np

a = np.array([1, 2, 3])           # 1D from list
b = np.array([[1, 2], [3, 4]])    # 2D from nested list
c = np.array([1, 2, 3], dtype=float)  # specify dtype

TEXT 📖 Display only

> **Output:** Run NumPy 2.x in your local Python environment to see the ndarray output. The Piston server does not have NumPy pre-installed — install it locally (`pip install numpy`) and compare. Actual values may vary by NumPy version and random seed.

Creation Methods Comparison

Method	Syntax	Input Source	Typical Use
`np.array`	`np.array(obj)`	Python list/tuple	Converting existing data
`np.arange`	`np.arange(start, stop, step)`	Numeric range	Integer/float sequences
`np.linspace`	`np.linspace(a, b, n)`	Start, stop, count	Precise element count
`np.zeros`	`np.zeros(shape)`	Shape	Initialize all zeros
`np.ones`	`np.ones(shape)`	Shape	Initialize all ones
`np.full`	`np.full(shape, val)`	Shape + fill value	Initialize to any value
`np.random.rand`	`np.random.rand(n)`	Element count	Random initialization
`np.fromfunction`	`np.fromfunction(fn, shape)`	Function + shape	Generate by formula

(2) Numeric Sequences

arange

PYTHON

np.arange(5)            # [0 1 2 3 4]
np.arange(1, 10, 2)     # [1 3 5 7 9]
np.arange(0, 1, 0.3)    # [0.  0.3 0.6 0.9]

TEXT 📖 Display only

> **Output:** Run NumPy 2.x in your local Python environment to see the ndarray output. The Piston server does not have NumPy pre-installed — install it locally (`pip install numpy`) and compare. Actual values may vary by NumPy version and random seed.

linspace

PYTHON

np.linspace(0, 1, 5)    # [0.   0.25 0.5  0.75 1.  ]
np.linspace(0, 2*np.pi, 100)  # 100 points from 0 to 2pi

TEXT 📖 Display only

> **Output:** Run NumPy 2.x in your local Python environment to see the ndarray output. The Piston server does not have NumPy pre-installed — install it locally (`pip install numpy`) and compare. Actual values may vary by NumPy version and random seed.

logspace

PYTHON

np.logspace(1, 3, 3)    # [  10.  100. 1000.]  -- 10^1, 10^2, 10^3
np.logspace(0, 2, 5)    # 5 points from 10^0 to 10^2

TEXT 📖 Display only

> **Output:** Run NumPy 2.x in your local Python environment to see the ndarray output. The Piston server does not have NumPy pre-installed — install it locally (`pip install numpy`) and compare. Actual values may vary by NumPy version and random seed.

arange vs linspace vs logspace

Feature	`arange`	`linspace`	`logspace`
Control	start, stop, step	start, stop, num	start_exp, stop_exp, num
Includes stop	No	Yes	Yes
Spacing	Linear	Linear	Logarithmic
Float precision	Can have errors	Exact	Exact
Typical use	Integer indices	Plotting samples	Frequency/log scales

(3) Pre-filled Arrays

PYTHON

np.zeros(3)             # [0. 0. 0.]
np.zeros((2, 3))        # 2x3 zero matrix
np.ones((2, 2))         # 2x2 ones
np.eye(3)               # 3x3 identity
np.full((2, 3), 7)      # 2x3 filled with 7
np.empty((2, 2))        # 2x2 uninitialized (NOT zeros!)

TEXT 📖 Display only

> **Output:** Run NumPy 2.x in your local Python environment to see the ndarray output. The Piston server does not have NumPy pre-installed — install it locally (`pip install numpy`) and compare. Actual values may vary by NumPy version and random seed.

zeros vs empty vs full

Function	Initial Value	Speed	Safety	Use Case
`zeros`	All zeros	Medium	Safe	Need a known initial value
`empty`	Undefined	Fast	Unsafe	Overwriting all elements immediately
`full`	Custom value	Medium	Safe	Need a non-zero initial value

(4) Random Number Creation

PYTHON

np.random.rand(3)           # 3 uniform [0, 1)
np.random.rand(2, 3)        # 2x3 uniform
np.random.randn(3)          # 3 standard normal
np.random.randint(0, 10, 5) # 5 ints in [0, 10)
np.random.seed(42)          # set seed for reproducibility

TEXT 📖 Display only

> **Output:** Run NumPy 2.x in your local Python environment to see the ndarray output. The Piston server does not have NumPy pre-installed — install it locally (`pip install numpy`) and compare. Actual values may vary by NumPy version and random seed.

Random Function Quick Reference

Function	Distribution	Range	Return Type
`rand(d0, d1, ...)`	Uniform	[0, 1)	float
`randn(d0, d1, ...)`	Standard normal	(-∞, +∞)	float
`randint(low, high, size)`	Uniform integer	[low, high)	int
`uniform(low, high, size)`	Uniform	[low, high)	float
`normal(loc, scale, size)`	Normal	(-∞, +∞)	float
`choice(a, size)`	Discrete uniform	Elements of a	depends

(5) Reading from Files

PYTHON

data = np.loadtxt('data.txt')                    # plain text
data = np.loadtxt('data.csv', delimiter=',')     # CSV
data = np.genfromtxt('data.csv', delimiter=',',  # handle missing
                     filling_values=0)

TEXT 📖 Display only

> **Output:** Run NumPy 2.x in your local Python environment to see the ndarray output. The Piston server does not have NumPy pre-installed — install it locally (`pip install numpy`) and compare. Actual values may vary by NumPy version and random seed.

fromfunction — generate arrays by coordinate function:

PYTHON

np.fromfunction(lambda i, j: i + j, (3, 3))
# [[0. 1. 2.]
#  [1. 2. 3.]
#  [2. 3. 4.]]

TEXT 📖 Display only

> **Output:** Run NumPy 2.x in your local Python environment to see the ndarray output. The Piston server does not have NumPy pre-installed — install it locally (`pip install numpy`) and compare. Actual values may vary by NumPy version and random seed.

▶ Example

TEXT 📖 Display only

> **Output:** Run NumPy 2.x in your local Python environment to see the ndarray output. The Piston server does not have NumPy pre-installed — install it locally (`pip install numpy`) and compare. Actual values may vary by NumPy version and random seed.

: arange vs linspace vs logspace (Difficulty ⭐)

PYTHON

import numpy as np

# arange: step-based
a = np.arange(0, 10, 2)       # [0 2 4 6 8]
print("arange:", a)

# linspace: num-based
b = np.linspace(0, 10, 6)     # 6 points: 0, 2, 4, 6, 8, 10
print("linspace:", b)

# logspace: logarithmic
c = np.logspace(1, 4, 4)      # [10, 100, 1000, 10000]
print("logspace:", c)

TEXT 📖 Display only

> **Output:** Run NumPy 2.x in your local Python environment to see the ndarray output. The Piston server does not have NumPy pre-installed — install it locally (`pip install numpy`) and compare. Actual values may vary by NumPy version and random seed.

⚠️ Note: The code below needs to run in a local Python environment.

▶ Example

TEXT 📖 Display only

> **Output:** Run NumPy 2.x in your local Python environment to see the ndarray output. The Piston server does not have NumPy pre-installed — install it locally (`pip install numpy`) and compare. Actual values may vary by NumPy version and random seed.

: zeros / ones / eye / full (Difficulty ⭐)

PYTHON

import numpy as np

z = np.zeros((2, 3))
print("zeros:\n", z)

o = np.ones((3, 2))
print("ones:\n", o)

e = np.eye(4)
print("eye:\n", e)

f = np.full((2, 3), 9.0)
print("full:\n", f)

TEXT 📖 Display only

> **Output:** Run NumPy 2.x in your local Python environment to see the ndarray output. The Piston server does not have NumPy pre-installed — install it locally (`pip install numpy`) and compare. Actual values may vary by NumPy version and random seed.

⚠️ Note: The code below needs to run in a local Python environment.

▶ Example

TEXT 📖 Display only

> **Output:** Run NumPy 2.x in your local Python environment to see the ndarray output. The Piston server does not have NumPy pre-installed — install it locally (`pip install numpy`) and compare. Actual values may vary by NumPy version and random seed.

: Nested list to 2D array (Difficulty ⭐)

PYTHON

import numpy as np

data = [[1, 2, 3],
        [4, 5, 6],
        [7, 8, 9]]

a = np.array(data)
print("shape:", a.shape)    # (3, 3)
print("dtype:", a.dtype)    # int64
print("ndim:", a.ndim)      # 2
print("array:\n", a)

TEXT 📖 Display only

> **Output:** Run NumPy 2.x in your local Python environment to see the ndarray output. The Piston server does not have NumPy pre-installed — install it locally (`pip install numpy`) and compare. Actual values may vary by NumPy version and random seed.

⚠️ Note: The code below needs to run in a local Python environment.

▶ Example

TEXT 📖 Display only

> **Output:** Run NumPy 2.x in your local Python environment to see the ndarray output. The Piston server does not have NumPy pre-installed — install it locally (`pip install numpy`) and compare. Actual values may vary by NumPy version and random seed.

: Random array generation (Difficulty ⭐)

PYTHON

import numpy as np

np.random.seed(42)

print("rand(3):", np.random.rand(3))
print("randn(3):", np.random.randn(3))
print("randint(0,10,5):", np.random.randint(0, 10, 5))
print("uniform(0,1,3):", np.random.uniform(0, 1, 3))
print("normal(0,1,3):", np.random.normal(0, 1, 3))

TEXT 📖 Display only

> **Output:** Run NumPy 2.x in your local Python environment to see the ndarray output. The Piston server does not have NumPy pre-installed — install it locally (`pip install numpy`) and compare. Actual values may vary by NumPy version and random seed.

⚠️ Note: The code below needs to run in a local Python environment.

4. Comparison Tables

(1) Creation Method by Use Case

Need	Method	Why
Convert existing data	`np.array(data)`	Direct conversion
Integer sequence	`np.arange(n)`	Lightweight
Precise floating sequence	`np.linspace(a, b, n)`	Exact count
Fixed value matrix	`np.full(shape, val)`	Any fill value
Identity matrix	`np.eye(n)`	Linear algebra
Random initialization	`np.random.rand(...)`	ML weights
Load from file	`np.loadtxt(file)`	External data

(2) arange vs Python range

Feature	`range`	`np.arange`
Returns	Lazy iterator	ndarray (materialized)
Memory	O(1)	O(n)
Float step	No	Yes
Speed (iteration)	Slower	Faster (vectorized)
Use case	Looping	Array creation

5. Principle Diagram

graph TB
    A[Python List] -->|np.array| B[ndarray]
    C[arange / linspace] --> B
    D[zeros / ones / eye] --> B
    E[random functions] --> B
    F[file: loadtxt] --> B
    B --> G[uniform dtype]
    B --> H[contiguous memory]
    B --> I[vectorized ops]

❓ FAQ

Q What's the difference between arange and linspace?

A arange takes a step size and doesn't include the stop value. linspace takes a count and includes both endpoints. Use arange for integer sequences, linspace when you need a specific number of points.

Q Is empty() really empty?

A No — it allocates the memory but doesn't initialize it. The "values" are whatever was in that memory before. It's faster but dangerous. Only use it when you're about to fill every element.

Q What's the modern way to generate random numbers?

A Use np.random.default_rng(seed) to create a Generator object, then call .random(), .normal(), .integers(), etc. The old np.random.seed() still works but is considered legacy.

Q Can I create an array from a generator expression?

A np.array(x for x in range(10)) doesn't work — it creates a 0D array containing the generator object. Use np.fromiter((x for x in range(10)), dtype=int) instead.

Q How do I handle missing data when loading CSV?

A Use np.genfromtxt('file.csv', delimiter=',', filling_values=0) to replace missing values with a default, or np.nan.

📖 Summary

np.array() converts Python lists to ndarrays; dtype is inferred automatically
arange(start, stop, step) creates evenly spaced sequences; linspace controls the count
zeros/ones/eye/full create pre-filled arrays; empty is faster but uninitialized
Random functions: rand (uniform), randn (normal), randint (integer), choice (sampling)
loadtxt and genfromtxt read arrays from text files; fromfunction generates by formula

📝 Exercises

Beginner (Difficulty ⭐): Create the following arrays: (a) integers 0 to 9, (b) 10 evenly spaced points from 0 to 1, (c) a 3x3 identity matrix, (d) a 2x4 array of all 7s.
Intermediate (Difficulty ⭐⭐): Use np.loadtxt to read a CSV file with 3 columns: name (string), age (int), salary (float). Use np.genfromtxt with dtype=None to handle the mixed types. Print the average salary.
Advanced (Difficulty ⭐⭐⭐): Benchmark creating 10 million random numbers using np.random.rand(10_000_000) vs a Python list comprehension. Record the time and memory difference. Then use np.random.default_rng() and compare the speed.

Web-Tutorial Tech Team

A team of developers maintaining programming tutorials. Each tutorial is written and reviewed by developers with expertise in that field. We work to keep our content accurate and reliable — if you spot an issue, please let us know.

About Us

NumPy: Creating Arrays

Creating Arrays

1. What You'll Learn

2. Story

3. Key Concepts

(1) Creating from Python Objects

(2) Numeric Sequences

arange

linspace

logspace

(3) Pre-filled Arrays

(4) Random Number Creation

(5) Reading from Files

▶ Example

▶ Example

▶ Example

▶ Example

4. Comparison Tables

(1) Creation Method by Use Case

(2) arange vs Python range

5. Principle Diagram

❓ FAQ

📖 Summary

📝 Exercises

🙏 帮我们做得更好